Simon Willison’s Weblog
Show HN: Simon Willison’s Weblog
Last verified:
What is Simon Willison’s Weblog?
Simon Willison’s Weblog is an open source extensible AI assistant plugin for Datasette that provides a conversational interface for asking questions about data stored in SQLite databases. It integrates LLM technology into Datasette, supporting hundreds of different tool-calling models from frontier vendors like OpenAI, Anthropic, and Google Gemini, as well as open weight models that can run on your own hardware.
Simon Willison’s Weblog pricing
Pricing model: Freemium
Datasette Agent is open source and free to use. The live demo at agent.datasette.io is free but requires GitHub sign-in. Running locally requires your own API key for cloud LLM providers (OpenAI, Anthropic, Google Gemini) or local hardware for open weight models. Alternative billing through OpenAI Codex is available via llm-openai-via-codex, which bills to your existing ChatGPT subscription. Datasette Agent will be rolled out for Datasette Cloud users (pricing details for Cloud not specified).
Simon Willison’s Weblog pros
- Open source and freely available
- Conversational interface for querying SQLite data
- Supports hundreds of tool-calling LLM models
- Works with OpenAI, Anthropic, and Google Gemini
- Supports open weight models for local deployment
- Extensible plugin architecture
- datasette-agent-charts creates visualizations with Observable Plot
- datasette-agent-openai-imagegen adds image generation with ChatGPT Images 2.0
- datasette-agent-sprites enables code execution in Fly Sprites sandbox
- Automatically generates correct SQL queries
- Export conversations to Markdown format
- Live demo available at agent.datasette.io
- Works with local models via LM Studio
- Can use OpenAI Codex billing through ChatGPT subscription
- Fast and inexpensive with Gemini 3.1 Flash-Lite
- Active development with Discord community support
- Easy to build custom plugins
- Coding agents like Codex and Claude Code can help build plugins
Simon Willison’s Weblog cons
- Currently in alpha release (0.1a3)
- Requires GitHub sign-in for the live demo
- Needs reliable tool-calling capability from LLM
- Model must be capable of producing valid SQLite queries
- Some plugins still in prototype stage
- API keys required for most cloud LLM providers
- Local model setup requires additional configuration
- Datasette Cloud integration still rolling out
Frequently asked questions about Simon Willison’s Weblog
What is Datasette Agent?
Datasette Agent is an open source plugin for Datasette that provides an extensible AI assistant for interacting with your SQLite databases. It integrates LLM into Datasette, providing a conversational interface for answering questions about your data in SQLite, enhanced by additional plugins to add visualizations and other custom tools.
Which LLM models does Datasette Agent support?
Datasette Agent supports hundreds of different tool-calling models from frontier vendors like OpenAI, Anthropic, and Google Gemini, through to open weight models that you can run on your own hardware. The live demo uses Gemini 3.1 Flash-Lite via the llm-gemini plugin.
How do I try the demo?
You can try the demo yourself at agent.datasette.io. This requires you to sign in with a GitHub account to help prevent abuse. The demo runs against example databases including the classic global-power-plants by WRI and a copy of the Datasette backup of Simon Willison's blog.
How do I run Datasette Agent locally?
The easiest way is using uv. For OpenAI, set your OPENAI_API_KEY environment variable and run: uvx --prerelease=allow --with datasette-agent datasette -s plugins.datasette-llm.default_model gpt-5.5 --internal internal.db --root legislators.db. For local models via LM Studio: uvx --prerelease=allow --with datasette-agent --with llm-lmstudio datasette --internal internal.db --root legislators.db -s plugins.datasette-llm.default_model lmstudio/qwen3.5-9b
What plugins are available for Datasette Agent?
Three plugins launched with the alpha: datasette-agent-charts adds charts powered by Observable Plot, datasette-agent-openai-imagegen adds image generation using ChatGPT Images 2.0, and datasette-agent-sprites provides tools for executing code in a Fly Sprites persistent sandbox. Several more plugins are in the pipeline.
Can I build my own plugins for Datasette Agent?
Yes, building plugins is described as really fun. The README includes detailed documentation, and coding agents like Codex and Claude Code are excellent at writing plugins - just point them at a checkout of the datasette-agent repo for reference and tell them what you want to build.
Does Datasette Agent work with local models?
Yes, Datasette Agent works with local models. The open weight models released in the past six months are increasingly able to handle the reliable tool calls and SQLite query generation that Datasette Agent needs. Examples include running gemma-4-26b-a4b in LM Studio on a Mac or Qwen 3.5 9B served via LM Studio.
What is the current release status?
The current release is version 0.1a3, which is the first alpha release. Today's alpha release is just the start, and the team is actively iterating on Datasette Agent and extra plugins for it.
How does Datasette Agent handle SQL queries?
Datasette Agent automatically generates SQL queries to answer natural language questions about your data. For example, asking 'when did Simon most recently see a pelican?' generates a SQL query that searches the blog_beat table for sighting entries containing 'pelican' and orders by creation date. It also provides 'View SQL query' buttons for both visible tables and collapsed SQL result tool calls.
Where can I get help or talk about Datasette Agent?
You can join the #datasette-agent Discord channel to talk about the project, learn more, and share what you are planning to build. The team is actively iterating and welcomes community participation.